Scale-dependent networked framing in a public crisis: Communication and semantic dynamics on X after the 2024 Trump assassination attempt

Abstract How do communication relations and semantic organization jointly change during a public crisis? Framing research has largely focused on textual frames or static semantic structures, leaving the temporal relationship between interaction networks and meaning configurations underexplored. We analyze 712,372 X posts about the July 2024 attempted assassination of Donald Trump. Across five event phases and six temporal resolutions (5–60 minutes), we construct directed mention networks and term-level semantic co-occurrence networks. Convergent cross mapping (CCM) assesses historical associations between communication and semantic structures, while transfer entropy (TE) tests short-term incremental predictive information. Results vary by phase and temporal scale. CCM associations were most prevalent at 5–10-minute resolutions during early and intermediate post-event periods, but generally did not coincide with robust one-step TE pathways. Robust TE appeared only during days 2–7 after the attack and at 5–10-minute resolutions. Four of five pathways supported in both raw and activity-adjusted series ran from communication measures to subsequent semantic properties; one ran in the opposite direction. During the lower-attention aftermath, CCM associations appeared only within a continuous jointly active segment, while no robust TE pathway emerged at the tested 20-minute resolution. These findings show that networked framing is temporally layered: historical association may coexist with intermittent, metric-specific predictive relationships, without demonstrating stable directional influence.